EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

An Analysis of Loss Functions for Heavily Imbalanced Lesion Segmentation

Mariano Cabezas iD, Yago Diez iD

Sensors · 20.03.2024 · 24 (6) · 1981

DOI: 10.3390/s24061981 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 20.03.2024. Verfasst von Mariano Cabezas und 1 weiteren Autor:innen. Beteiligte Einrichtung: Brain and Mind Centre, The University of Sydney, Camperdown, NSW 2050, Australia. Hauptthemen laut MEDLINE: Neural Networks, Computer, Magnetic Resonance Imaging. Die Arbeit wurde bisher 13-mal zitiert (Crossref) und nennt 28 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Heavily imbalanced datasets are common in lesion segmentation. Specifically, the lesions usually comprise less than 5% of the whole image volume when dealing with brain MRI. A common solution when training with a limited dataset is the use of specific loss functions that rebalance the effect of background and foreground voxels. These approaches are usually evaluated running a single cross-validation split without taking into account other possible random aspects that might affect the true improvement of the final metric (i.e., random weight initialisation or random shuffling). Furthermore, the evolution of the effect of the loss on the heavily imbalanced class is usually not analysed during the training phase. In this work, we present an analysis of different common loss metrics during training on public datasets dealing with brain lesion segmentation in heavy imbalanced datasets. In order to limit the effect of hyperparameter tuning and architecture, we chose a 3D Unet architecture due to its ability to provide good performance on different segmentation applications. We evaluated this framework on two public datasets and we observed that weighted losses have a similar performance on average, even though heavily weighting the gradient of the foreground class gives better performance in terms of true positive segmentation.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Neural Networks, ComputerMagnetic Resonance ImagingNeuroimagingImage Processing, Computer-Assisted

Stichwörter der Autor:innen

brainimbalanced datasetlesion segmentationloss functionsmagnetic resonance imagingmedical imaging sensors

Autor:innen und Institutionen

  • Brain and Mind Centre, The University of Sydney, Camperdown, NSW 2050, Australia.
  • Faculty of Science, Yamagata University, Yamagata 990-8560, Japan.

ORCID-Profile: Mariano Cabezas, Yago Diez

Zitieren

APA

Cabezas, M., & Diez, Y. (2024). An Analysis of Loss Functions for Heavily Imbalanced Lesion Segmentation. Sensors, 24(6), 1981. https://doi.org/10.3390/s24061981

Vancouver

Cabezas M, Diez Y. An Analysis of Loss Functions for Heavily Imbalanced Lesion Segmentation. Sensors. 2024;24(6):1981. doi:10.3390/s24061981

BibTeX
@article{Cabezas2024,
  author = {Cabezas, Mariano and Diez, Yago},
  title = {An Analysis of Loss Functions for Heavily Imbalanced Lesion Segmentation},
  journal = {Sensors},
  year = {2024},
  volume = {24},
  number = {6},
  pages = {1981},
  doi = {10.3390/s24061981}
}
Kostenlos starten →